Quick start - AI Tour Meeting

Prerequisites

The following must be installed on your machine:

Docker (official Docker installation guide)
Make (apt install make)

1. Clone the repository

Clone the repository and navigate into the project directory.

git clone https://github.com/ntt-dkiku/ai-tour-meeting.git
cd ai-tour-meeting

2. Set your API keys (optional)

Create your .env file and add your API keys.

cp docker/.env.example docker/.env

Tip

You can also set API keys from the GUI (Settings in the sidebar), so this step can be skipped. However, in that case, you will need to enter the keys every time you open the GUI. If you prefer not to store your API keys in plain text in the .env file, this may be a better option.

3. Deploy AI Tour Meeting

Deploy AI Tour Meeting with the following command:

make up

After the deployment, access localhost:3000 in your browser, and you can try the AI Tour Meeting GUI! (You can also use make up OPEN_GUI=1 to open the browser automatically.)

Tip

If you are working on a remote server, don't forget to set up port forwarding, e.g., ssh <remote-server-name> -L 3000:localhost:3000 -L 8080:localhost:8080. The backend port defaults to 8080 and can be changed via the BACKEND_PORT environment variable (e.g., BACKEND_PORT=9090 make up).

3.1. Local LLMs

Ollama

If you want to use local LLMs via Ollama, add the OLLAMA option: OLLAMA=cpu for CPU mode, or OLLAMA=gpu for GPU mode (NVIDIA GPUs only).

# CPU mode
make up OLLAMA=cpu
# Use all GPUs in a single ollama server
make up OLLAMA=gpu
# Specify GPUs used in a single ollama server
make up OLLAMA=gpu:0,1
# Specify GPUs used in separated ollama servers
make up OLLAMA="gpu:0,1 gpu:2,3"

vLLM (GPU only)

You can also use vLLM to serve local models. Add the VLLM option with the model name. An NVIDIA GPU is required.

# Single model on single GPU
make up VLLM=Qwen/Qwen3-8B
# Single model on multiple GPUs
make up VLLM="Qwen/Qwen3-32B:gpu=0,1"
# Multiple models on separated GPUs
make up VLLM="Qwen/Qwen3-8B:gpu=0 openai/gpt-oss-20b:gpu=1"
# Multiple models on multiple GPUs
make up VLLM="Qwen/Qwen3-32B:gpu=0,1 openai/gpt-oss-120b:gpu=2,3"

Note

For gated models (e.g., Llama), set HF_TOKEN in docker/.env.

4. Use AI Tour Meeting

AI Tour Meeting provides three interfaces. Pick the one that fits your use case:

GUI

Create and run meetings from the browser app.

CLI

Run meetings from the terminal.

Python API

Script and customize meetings in Python.